DeepSource Launches Autonomous AI Agents to Combat Rising Cybersecurity Threats

According to a recent report, the global cost of cybercrime is projected to reach a staggering $10.5 trillion by 2025. Despite efforts by organizations and leaders, securing expanding digital infrastructures remains a challenge, especially with cloud and off-prem technologies. Research indicates that breaches often go undetected for an average of 212 days, with an additional 75 days required to contain them, posing significant risks to sensitive data.

DeepSource's AI-Driven Solution

DeepSource, a Y Combinator-backed startup, has unveiled three autonomous AI agents designed to enhance cybersecurity:

1. False-positive Triage Agent

This agent evaluates security issues in code, suppressing invalid alerts based on repository context and real-world threat intelligence.

2. CVE Prioritization Agent

Automatically triages open-source vulnerabilities, reducing the manual workload for AppSec teams.

3. Autofix™ AI Autopilot

Leverages DeepSource's existing Autofix™ AI to autonomously create pull-requests with security fixes, learning from developer behavior.

Key Benefits

• Autonomy: Agents operate 100% autonomously.
• Time Savings: Saves ~5 hours per developer weekly.
• Contextual Understanding: Agents reason based on project context and team goals.

Industry Context

CEO Sanket Saurav emphasized the need for evolved tooling to secure the surge in AI-generated code, noting that 1 in 4 Y Combinator startups rely on AI for 95% of their code. However, nearly half of AI-generated code contains exploitable bugs. DeepSource's co-founder, Jai Pradeesh, clarified that their AI agents are goal-based, leveraging hundreds of signals for autonomous action, with transparent reasoning visible to users.

Open-Source Contribution

In February, DeepSource released Globstar, an open-source project offering cutting-edge code security tools to the AppSec community.

"Real end users will be impacted if companies don’t evolve their tooling to ensure they’re securing this exponentially higher volume of code," warned Saurav.